Report as of 2026-10-30 PT · Source: Metix AI
Notes & method
01 They sell software, yet over 4 in 10 face customers directly
Scope: current employees of the 12 companies, matched to each verified company page, excluding interns, advisors, investors and board members; data as of 2026-04.
Leaving out the outsourcer Crescendo, R&D (engineering, AI, product, design) is 31.0% of the 11 product companies' 3,829 people and researcher / scientist titles just 2.1%; roles that face customers directly (agent delivery, solutions and implementation, success and support, sales) add up to 43.2%, of which sales is 18.9%.
At the 10 product companies other than Crescendo and Gradient Labs, the customer-facing technical layer is 20.7%–29.5% of staff. Agent delivery makes up 80% and 95% of it at Sierra and Lorikeet, about half at Decagon and PolyAI, about 40% at Cresta and Parloa, and no more than 6% at Intercom, Ada, Cognigy and Forethought, which rely on solutions, implementation and support.
The post-2023 generation (Sierra, Decagon, Lorikeet): flat titles, about 70% hired in the past year. Contact-center and voice AI (Cresta, Parloa, PolyAI, Cognigy): customer teams far more US-weighted than R&D. Earlier platforms (Intercom, Ada, Forethought): mature ladders, long tenure. AI-native outsourcing (Crescendo): 55.1% frontline CX operations.
Grouping everyone by join date (leavers included, Crescendo excluded): among people hired in the past 12 months, solutions and implementation is 10.8% (earlier hires 5.5%) and rises at all 11 companies; software engineering is 15.0% (18.3%) and falls at 9 of 11; sales is 22.3% (19.7%) and rises at 8 of 11. Agent delivery rises from 2.8% to 11.6%, with Sierra accounting for 83 of 195.
Only 5.1% had a CX or contact-center software company (including these 12) as their most recent previous employer, and just 22 people have worked at another of the 12. The most common destinations of Intercom's 1,409 employed alumni are HubSpot, Google, Meta and Salesforce (5.9% combined); 46.5% are at startups and scale-ups, and 7 work at the other 11.
Only 1.9% of Sierra's non-sales titles carry a level such as Senior or Staff, against 39.0% at Intercom and 42.0% at Parloa; Sierra titles its account executives Sales Director (53 people). Read each company's naming before searching by title or benchmarking levels.
02The 12 companies: size, hubs and two dividing lines
The first dividing line is time: everyone at Sierra, Decagon and Lorikeet joined in 2023 or later, about 70% in the past year, and they carry the heaviest agent-delivery teams. The second is product: Cresta, Parloa, PolyAI and Cognigy build for contact centers and voice, Intercom is the incumbent help desk going AI, and Crescendo is AI-native outsourcing with its own frontline teams.
The 12 at a glance
Visible = current employees mapped here; page = headcount on the company page; * = under 100 people.
Intercom
- Visible / page
- 1,181 / 1,982
- Largest city hub
- Dublin 37.9%
- R&D
- 31.2%
- Delivery layer
- 21.7%
- Joined <12 mo.
- 26.4%
Crescendo
- Visible / page
- 791 / 2,270
- Largest city hub
- Honduras 36.4%
- R&D
- 5.2%
- Delivery layer
- 4.4%
- Joined <12 mo.
- 31.0%
Sierra
- Visible / page
- 508 / 721
- Largest city hub
- SF Bay Area 57.9%
- R&D
- 26.6%
- Delivery layer
- 25.8%
- Joined <12 mo.
- 67.5%
Cresta
- Visible / page
- 473 / 649
- Largest city hub
- SF Bay Area 20.9%
- R&D
- 32.1%
- Delivery layer
- 25.8%
- Joined <12 mo.
- 49.0%
Parloa
- Visible / page
- 376 / 433
- Largest city hub
- Berlin 46.3%
- R&D
- 28.2%
- Delivery layer
- 20.7%
- Joined <12 mo.
- 40.4%
Decagon
- Visible / page
- 294 / 388
- Largest city hub
- SF Bay Area 64.3%
- R&D
- 30.6%
- Delivery layer
- 26.2%
- Joined <12 mo.
- 73.8%
PolyAI
- Visible / page
- 275 / 360
- Largest city hub
- London 29.8%
- R&D
- 32.0%
- Delivery layer
- 29.5%
- Joined <12 mo.
- 34.9%
Ada
- Visible / page
- 269 / 677
- Largest city hub
- Toronto 38.7%
- R&D
- 35.7%
- Delivery layer
- 25.3%
- Joined <12 mo.
- 28.3%
Cognigy
- Visible / page
- 260 / 319
- Largest city hub
- Düsseldorf 20.4%
- R&D
- 29.2%
- Delivery layer
- 28.5%
- Joined <12 mo.
- 29.6%
Forethought*
- Visible / page
- 92 / 105
- Largest city hub
- SF Bay Area 19.6%
- R&D
- 38.0%
- Delivery layer
- 23.9%
- Joined <12 mo.
- 23.9%
Lorikeet*
- Visible / page
- 65 / 90
- Largest city hub
- Sydney 38.5%
- R&D
- 33.8%
- Delivery layer
- 29.2%
- Joined <12 mo.
- 69.2%
Gradient Labs*
- Visible / page
- 36 / 45
- Largest city hub
- London 55.6%
- R&D
- 47.2%
- Delivery layer
- 8.3%
- Joined <12 mo.
- 52.8%
Hiring pace × agent-delivery share
Company: hired in the past 12 months · agent-delivery share
- Lorikeet 69.2% · 27.7%
- Sierra 67.5% · 20.7%
- PolyAI 34.9% · 15.6%
- Decagon 73.8% · 12.9%
- Cresta 49.0% · 10.8%
- Parloa 40.4% · 8.5%
- Intercom 26.4% · 1.4%
- Cognigy 29.6% · 0.8%
- Crescendo 31.0% · 0.6%
- Forethought 23.9% · 0.0%
- Ada 28.3% · 0.0%
- Gradient Labs 52.8% · 0.0%
03Org structure: R&D about 3 in 10; sales and engineering largest
Each person is assigned to one of 20 functions in 7 groups by current title. R&D is 26.6%–38.0% at the 10 product companies and nearly half at Gradient Labs (47.2%); more than half of Crescendo is frontline CX operations. Sales (787 people) and software engineering (741) are the two largest functions.
Seven function groups per company
Sorted by R&D share; numbers are headcount.
Headcount by 20 functions (all 12)
The 66 people (1.4%) whose function could not be identified are not shown; 2 of them have blank titles.
04R&D and research: few ML and research titles
Software engineering dominates R&D; AI / ML and research roles are 11.2% of engineering, and researcher / scientist titles 1.7% of all staff at the 12. The exception is Gradient Labs, where research is 16.7% of 36 people. Intercom has the largest R&D team (369 people), with AI / ML at 16.6% of engineering.
Inside R&D
Sorted by AI / ML share; numbers are R&D headcount.
Researcher and scientist titles, share of all staff
05The delivery layer: same size, three make-ups
AI customer service only works once it is wired into a customer's systems, processes and knowledge, so every company carries a customer-facing technical layer (agent delivery, solutions and implementation, customer success, support). It is 20.7%–29.5% of staff at the 10 product companies, with very different shares of agent delivery: most of it at Sierra and Lorikeet, about half at Decagon and PolyAI, about 40% at Cresta and Parloa; Intercom, Ada, Cognigy and Forethought rely on solutions, implementation and support, and nearly half of Intercom's layer is support (10.2% of all staff). Crescendo is an outsourcer and Gradient Labs is 47.2% R&D; both have a thin layer.
Customer-facing technical layer: size and make-up
Bar length is the share of all staff (longest bar = full width), total next to the name; sorted by agent-delivery share.
06Titles and levels: flat newcomers, laddered incumbents
Sales titles follow each company's own convention (Sierra calls account executives Sales Directors), so levels are read on non-sales roles only. The new generation barely puts levels in titles: 1.9% at Sierra carry Senior, Staff, Principal and the like, 0.0% at Gradient Labs; Intercom, Ada, Forethought, Cognigy, Parloa and PolyAI sit at 31.7%–42.0%. Sierra has the fewest manager titles: 1 manager or above for every 16.9 non-managers.
Titles carrying a level (excl. sales)
Non-managers per manager or above
Crescendo's frontline team leads count as managers; frontline staff are under-represented in visible profiles, so its ratio is understated.
The two most common sales titles
07Hiring momentum: who is adding people, and where
In the 12 months to 2026-04, 39.7% of current employees were hired; Decagon (73.8%), Lorikeet (69.2%) and Sierra (67.5%) lead, while Forethought, Intercom, Ada, Cognigy are under 30%. Where new people go says more than how many: the share in solutions and implementation rises at all 11 product companies, software engineering falls at 9 of 11, and 83 of the 195 new agent-delivery hires are at Sierra.
Current employees by year joined
Counts only people still employed, so earlier years are understated; the pale bar is the first 4 months of 2026.
Where new hires go (excl. Crescendo)
Current and former employees grouped by join date; leavers are classified by their last title at the company.
08Geography: R&D stays home, customer teams lean to the US
7 companies keep under 15% of R&D staff in the US (Intercom, Ada, Lorikeet, Parloa, Cognigy, PolyAI, Gradient Labs): none above 11.7%, while their customer-facing teams (agent delivery, solutions, success, support, sales and marketing) are far more US-weighted, 38.5%–58.0% except Ada. Cresta has 61.2% of R&D outside the US and 91.0% of customer-facing staff in it.
US share of R&D vs customer-facing teams
Sorted by the US share of R&D.
| Company | Main R&D country | R&D in US % | Customer-facing in US % |
|---|---|---|---|
| Gradient Labs* | UK 100.0% | 0.0 | 38.5 |
| Cognigy | Germany 90.8% | 3.9 | 38.6 |
| Parloa | Germany 82.1% | 5.7 | 40.7 |
| PolyAI | UK 78.4% | 5.7 | 58.0 |
| Ada | Canada 89.6% | 8.3 | 12.8 |
| Lorikeet* | Australia 90.9% | 9.1 | 46.2 |
| Intercom | Ireland 49.9% | 11.7 | 51.0 |
| Crescendo | Honduras 36.6% | 29.3 | 48.0 |
| Cresta | US 38.8% | 38.8 | 91.0 |
| Forethought* | US 71.4% | 71.4 | 75.5 |
| Sierra | US 94.8% | 94.8 | 82.3 |
| Decagon | US 98.9% | 98.9 | 96.6 |
09Where people come from and go: in from outside, out to outside
We read each person's previous employer before joining, and where each company's former employees work now. Several companies carry a clear lineage: 49.7% of Crescendo came from outsourcers (mostly PartnerHero, which it acquired), 34.3% of Gradient Labs from Monzo; Cognigy and PolyAI draw from CX and contact-center software at 15.2% and 13.4%, about 3 times the average (5.1%); Sierra's top three sources are Google, Stripe and Figma (8.8% combined). Overall, people come from outside the category: only 22 have worked at another of the 12.
Type of previous employer
Sorted by Big Tech share; “large” and “startups” split other employers by headcount.
Top three previous employers
| Company | Previous employer (share of known sources) |
|---|---|
| Intercom | Salesforce 2.2%, Meta 2.0%, Zendesk 1.8% |
| Crescendo | PartnerHero 38.8%, Concentrix 2.3%, STARTEK 1.7% |
| Sierra | Google 3.2%, Stripe 3.2%, Figma 2.4% |
| Cresta | Google 5.1%, Amazon 2.6%, Observe.AI 2.6% |
| Parloa | HelloFresh 1.9%, Salesforce 1.6%, Wayfair 1.6% |
| Decagon | Scale AI 4.5%, Bain & Company 4.1%, Salesforce 2.4% |
| PolyAI | Dialpad 3.4%, Amazon 2.7%, Replicant 1.9% |
| Ada | Qualtrics 1.9%, Deloitte Canada 1.6%, Shopify 1.6% |
| Cognigy | RingCentral 3.1%, Avaya 2.3%, Genesys 2.0% |
| Forethought* | Cisco 2.4%, Meta 2.4%, Darktrace 2.4% |
| Lorikeet* | Dovetail 7.8%, Google 6.2%, Bain & Company 4.7% |
| Gradient Labs* | Monzo Bank 34.3%, Pleo 5.7%, Cordless 5.7% |
Where alumni work now
Companies with 100+ employed alumni. Zendesk (Forethought) and NiCE (Cognigy) reflect acquisitions.
| Company | Employed alumni | Top destinations | At the other 11 |
|---|---|---|---|
| Intercom | 1,409 | HubSpot 29, Google 20, Meta 17 | 7 |
| Ada | 536 | Shopify 11, Salesforce 8, Zendesk 6 | 2 |
| Forethought* | 207 | Zendesk 16, Salesforce 4, Fieldguide 3 | 2 |
| Cresta | 183 | Google 10, VAST Data 3, Simile 3 | 0 |
| Parloa | 137 | Salesforce 4, IDnow 3, Choco 3 | 3 |
| PolyAI | 137 | Amazon 3 | 4 |
| Cognigy | 115 | Kore.ai 6, NiCE 3, Synthflow AI 3 | 2 |
10Education and experience: mostly bachelor's, mostly senior
Among people with education records, 2.7% hold a PhD and 52.2% a bachelor's as their highest degree. Schools follow the hubs: Stanford and Berkeley at Sierra and Decagon, Waterloo and Toronto at Cresta, Dublin universities at Intercom, Cambridge and Imperial at PolyAI, Honduran universities at Crescendo. Median career length is 11.9 years; Decagon is youngest (8.8), Cognigy most senior (14.5).
Highest degree
Covers the 84.5% with education records; sorted by PhD plus master's share.
Top schools per company (5+ people)
| Company | Schools |
|---|---|
| Intercom | University College Dublin · Trinity College Dublin · TU Dublin |
| Crescendo | UNITEC · CEUTEC · UNAH |
| Sierra | Stanford University · UC Berkeley · University of Southern California |
| Cresta | University of Waterloo · University of Toronto · UC Berkeley |
| Parloa | Technische Universität Berlin · Freie Universität Berlin · Humboldt-Universität zu Berlin |
| Decagon | UC Berkeley · Stanford University · Harvard University |
| PolyAI | Imperial College London · University of Cambridge · UCL |
| Ada | Queen's University · University of Toronto · Western University |
| Cognigy | Dispersed (no school with 5+) |
| Forethought* | Dispersed (no school with 5+) |
| Lorikeet* | UNSW · University of Sydney |
| Gradient Labs* | Dispersed (no school with 5+) |
Median career length (years)
11Key people: engineering and customer-facing leaders at the 12
For each company, 1 engineering & AI leader and 1 customer-facing leader (agent delivery, solutions, success, support), the most senior by title (chief > SVP > VP > senior director / head > director). Full org charts and lists unlock on the Metix AI platform.
Intercom
Crescendo
Sierra
Cresta
Parloa
Decagon
PolyAI
Ada
Cognigy
Forethought
Lorikeet
Gradient Labs
12Playbook
The data above, translated into moves for three readers.
Hire for the delivery layer; search by each house's names
- Plan the customer-facing technical layer as a core function: it is 20.7%–29.5% of staff at the 10 product companies, and 28.2% of people hired in the past 12 months against 24.2% for R&D (excl. Crescendo).
- Search with each house's names: Agent Engineer, Agent Product Manager, Forward Deployed Engineer, Solutions Consultant; a Sierra Sales Director is an account executive.
- Look beyond rivals: only 22 people have moved between the 12; at the 11 product companies the most recent previous employer was a startup or scale-up for 43.9% and Big Tech for 12.6%.
- Translate levels before benchmarking: flat-title companies (Sierra 1.9% leveled titles) do not map one-to-one onto Intercom (39.0%).
Use the workforce for vendor diligence
- Ask who will deliver: agent delivery is 27.7% at Lorikeet, 20.7% at Sierra, 12.9% at Decagon; Ada, Cognigy and Forethought rely on solutions and implementation teams.
- Check delivery tenure: 77.9% of Decagon's and 78.6% of Sierra's customer-facing technical layer joined in the past 12 months.
- Check time zones: Parloa, PolyAI and Cognigy build in Europe, with 38.6%–58.0% of customer-facing staff in the US.
- R&D depth is not AI research: researcher and scientist titles are 1.7% across the 12; only Gradient Labs (16.7%) is clearly higher.
Read org design as a leading indicator
- Read growth from hiring mix: solutions and sales rise among people hired in the past 12 months; software engineering falls from 18.3% to 15.0%.
- Recruiting teams signal expansion: 7.7% of Sierra and 9.5% of Decagon work in recruiting, against 3.7% overall.
- Acquisitions bring organisations, not just products: Crescendo's frontline came with PartnerHero, Cognigy is now part of NICE and Forethought part of Zendesk.
- Track where alumni go: the most common destinations of Intercom's 1,409 employed alumni are HubSpot, Google, Meta and Salesforce; 7 work at the other 11.
13Definitions
The four scopes behind the numbers.
Questions this report answers
- Which companies and how many people does the AI customer-service talent map cover?
- 4,620 current employees of 12 companies: Sierra, Decagon, Cresta, Intercom, Forethought, Crescendo, Ada, Lorikeet, Parloa, Cognigy, PolyAI and Gradient Labs. Each company is matched by its verified company page and a dated current role; interns, students, advisors, investors and board members are excluded; data as of 2026-04.
- How are AI customer-service companies staffed?
- Outside the outsourcer Crescendo, R&D is 31.0% of headcount and research 2.1%; customer-facing roles add up to 43.2%. The 10 product companies each carry a customer-facing technical layer of 20.7%–29.5%: mostly agent delivery at Sierra and Lorikeet, about half at Decagon, about a third at Cresta, Parloa and PolyAI, and solutions, implementation and support at Intercom, Ada, Cognigy and Forethought.
- Where does AI customer-service talent come from and go?
- Mostly from outside the category: only 5.1% previously worked at CX or contact-center software companies and 22 of 4,620 current employees have worked at another of the 12. The most common destinations of Intercom's 1,409 employed alumni are HubSpot, Google, Meta and Salesforce; 7 work at the other 11.
- How should this report be cited?
- Metix AI Talent Intelligence, 2026-10-30. The AI Customer-Service Talent Map: one market, four org models | Metix AI. https://metix.ai/reports/mapping/ai-customer-service-talent-map-2026
Want a full org chart for any of these companies?
Filter current employees and alumni of the 12 by function, level, hub and source — or build the same x-ray for any target company.
Start a Mira trial